The Born-Oppenheimer Approximation

Quantum mechanics hands molecular physics an equation it cannot solve. A molecule is a swarm of electrons and nuclei bound by Coulomb forces, obeying a single many-body Schrödinger equation in which every coordinate is entangled with every other — intractable for anything past $\mathrm{H}_2^+$ . The Born-Oppenheimer approximation rescues the whole subject with one fact from mechanics: the nuclei are thousands of times heavier than the electrons. On the timescale an electron needs to circle the molecule, the nuclei barely stir, so the electrons see a frozen framework of point charges and relax instantly into a stationary state fitted to it. The nuclei, in turn, feel only the smeared-out, time-averaged electron cloud — an effective potential. That single observation splits one impossible problem into two solvable ones. First, clamp the nuclei at a geometry $R$ and solve the electronic problem for an energy $E_n^{\mathrm{el}}(R)$ ; sweeping $R$ traces a potential energy surface. Second, let the nuclei move on that surface to get vibration and rotation. The molecular energy then falls into the ladder $E_{\mathrm{el}}\gg E_{\mathrm{vib}}\gg E_{\mathrm{rot}}$ , in the ratio $1:(m_e/M)^{1/2}:(m_e/M)$ — visible, infrared, microwave. Almost every model in molecular spectroscopy is a child of this approximation: bonding curves, molecular orbitals, harmonic and Morse oscillators, the rigid rotor, the Franck-Condon principle. Knowing where it holds — and where, at a conical intersection, it fails catastrophically — is the price of reading a molecular spectrum.

Key takeaways

The Born-Oppenheimer (BO) approximation exploits the huge nuclear-to-electron mass ratio ( $M/m_e\gtrsim1836$ ) to separate fast electronic from slow nuclear motion.
At each frozen geometry $R$ one solves the clamped-nuclei electronic equation; its eigenvalue $E_n^{\mathrm{el}}(R)$ becomes a potential energy surface governing the nuclei.
The total wavefunction factorizes, $\Psi(r,R)\approx\psi_n^{\mathrm{el}}(r;R)\,\chi_{n\nu}(R)$ , sorting structure into electronic, vibrational, and rotational tiers in the ratio $1:(m_e/M)^{1/2}:(m_e/M)$ .
The neglected nonadiabatic couplings carry $1/(E_m^{\mathrm{el}}-E_n^{\mathrm{el}})$ , so BO is excellent when electronic states are well separated and fails at conical intersections.
BO underpins nearly all of molecular spectroscopy — potential curves, harmonic and Morse vibration, the rigid rotor, and the Franck-Condon principle all presuppose it.
Key idea: freeze the nuclei, solve the electronic problem to get a potential energy surface $E_n^{\mathrm{el}}(R)$ , then let the nuclei move on that surface; the molecular wavefunction factorizes into an electronic part depending parametrically on $R$ times a nuclear part.

Where this lives today

The BO surface is the input to essentially all of computational chemistry. When a quantum-chemistry package reports a bond length, vibrational spectrum, or reaction barrier, it has solved equation (2) at many geometries to build $E_n^{\mathrm{el}}(R)$ and then moved the nuclei — classically in a trajectory, or quantum-mechanically for spectroscopy — on that surface. Density functional theory, coupled cluster, and the force fields that simulate proteins all live within the BO picture — "geometry optimization" just means finding a minimum of a Born-Oppenheimer surface. The frontier is where BO fails on purpose. Femtosecond and attosecond experiments now watch molecules pass through conical intersections in real time — the funnels that let a photoexcited molecule dump electronic energy into nuclear motion within tens of femtoseconds. This drives the first step of vision (the retinal chromophore twisting after absorbing a photon), the photostability of DNA (which sheds ultraviolet energy before bonds break), and the charge separation that begins photosynthesis. Modeling any of these requires non-adiabatic dynamics — surface-hopping and multi-configuration methods that reinstate the couplings $F_{nm}^{(1)}$ exactly where the single-surface picture surrenders. BO is thus both the workhorse of molecular science and the baseline whose deliberate breakdown defines ultrafast photochemistry.

The physical picture

Picture the nuclei as slow, heavy planets and the electrons as fast, light satellites. Over one electronic orbit the nuclei barely drift, so the electrons see a static framework and settle into a stationary state adapted to it. The operative word is adiabatic: if the nuclei move slowly enough the electronic state deforms continuously with them and never jumps to another surface — the quantum adiabatic theorem, with the nuclear coordinates as slowly varying parameters. The consequence is a two-step recipe. Solve the electrons at frozen $R$ , collect $E_n^{\mathrm{el}}(R)$ for every geometry, and you have a potential energy curve (diatomic) or surface (polyatomic) — a real, computable object with a minimum at the equilibrium bond length $R_e$ and depth equal to the dissociation energy $D_e$ . The nuclei then live in that well: its curvature sets the vibrational frequency, its width the rotational constant. The first figure shows this surface for $\mathrm{H}_2$ , with the vibrational levels the nuclei occupy.

Figure

The Born-Oppenheimer potential curve of H₂ — solving the clamped-nuclei electronic problem at each R traces E_el(R), a Morse-like well with minimum at R_e ≈ 0.74 Å and depth D_e ≈ 4.75 eV; the nuclei move on this surface, their low vibrational quanta spaced by ħω ≈ 0.55 eV

The second thing the picture buys is the energy hierarchy, which follows from the mass ratio alone (Derivation 1): vibrational quanta are smaller than electronic ones by $\sqrt{m_e/M}$ and rotations by a full power of $m_e/M$ , giving $1:10^{-2}:10^{-4}$ — exactly the visible / infrared / microwave split of molecular spectra. But the same approximation carries a fault line: where two surfaces approach, the electrons can no longer decide which state to occupy as the nuclei move, and the discarded couplings blow up. The second figure shows that breakdown — an avoided crossing where BO fails.

Figure

Where BO fails — two electronic surfaces that would cross (dashed diabatic curves) instead avoid, opening a minimum gap ΔE_min ≈ 0.5 eV; the nonadiabatic coupling F_nm ∝ 1/ΔE peaks in the shaded region and mixes the surfaces

The core equations

Everything begins with the exact molecular Hamiltonian, the object BO tames. It contains two kinetic terms and three Coulomb terms, with the electron-nucleus attraction $V_{eN}(r,R)$ the sole term entangling electrons with nuclei.

\hat H = \hat T_N + \hat T_e + V_{ee}(r) + V_{NN}(R) + V_{eN}(r,R) \qquad (1)

where: $\hat T_N=-\sum_A\frac{\hbar^2}{2M_A}\nabla_A^2$ — nuclear kinetic energy, suppressed by the large masses $M_A$ ; $\hat T_e=-\sum_i\frac{\hbar^2}{2m_e}\nabla_i^2$ — electronic kinetic energy; $V_{ee},V_{NN},V_{eN}$ — electron-electron, nucleus-nucleus, and electron-nucleus Coulomb interactions; $r,R$ — collective electronic and nuclear coordinates. Reading it: the difficulty is that $V_{eN}$ ties $r$ to $R$ , blocking a naive factorization — until the small size of $\hat T_N$ is exploited. The first step is to freeze the nuclei. Deleting $\hat T_N$ leaves an electronic Hamiltonian in which $R$ appears only inside multiplicative potentials, so it is a parameter, not a dynamical variable. Diagonalizing it at each $R$ is the clamped-nuclei problem, and its eigenvalue is the potential energy surface.

\hat H_{\mathrm{el}}(R)\,\psi_n^{\mathrm{el}}(r;R) = E_n^{\mathrm{el}}(R)\,\psi_n^{\mathrm{el}}(r;R) \qquad (2)

where: $\hat H_{\mathrm{el}}=\hat T_e+V_{ee}+V_{eN}+V_{NN}$ , at fixed $R$ ; $E_n^{\mathrm{el}}(R)$ — the $n$ -th electronic eigenvalue, whose $R$ -dependence is the potential energy surface; $\psi_n^{\mathrm{el}}(r;R)$ — electronic wavefunction depending parametrically on $R$ (the semicolon flags $R$ as a label, not a coordinate). Reading it: solve once per geometry and $E_n^{\mathrm{el}}(R)$ is the curve of the first figure. With the surfaces in hand, BO writes the full wavefunction as a single product of one electronic state and a nuclear amplitude on that surface. This is an ansatz, not an identity — the approximation lives here.

\Psi(r,R) \approx \psi_n^{\mathrm{el}}(r;R)\,\chi_{n\nu}(R) \qquad (3)

where: $\chi_{n\nu}(R)$ — nuclear (vibrational-rotational) wavefunction on electronic surface $n$ ; $\nu$ — nuclear quantum numbers (vibrational $\nu$ , rotational $J$ ). Reading it: the electrons are slaved to the instantaneous geometry; the nuclei carry the remaining freedom on a single surface. Substituting the ansatz and keeping the leading term gives the nuclei their own Schrödinger equation, with the electronic surface as the potential — the equation every vibration-rotation model solves.

\left[\hat T_N + E_n^{\mathrm{el}}(R)\right]\chi_{n\nu}(R) = E_{n\nu}\,\chi_{n\nu}(R) \qquad (4)

where: $E_n^{\mathrm{el}}(R)$ — the electronic surface, now a potential well for the nuclei; $E_{n\nu}$ — total molecular energy (electronic plus vibrational plus rotational). Reading it: the electrons are integrated out, leaving a clean nuclear problem in a known potential. Expanding $E_n^{\mathrm{el}}(R)$ about its minimum and adding rotation gives the three-tier energy ladder — the skeleton of every molecular spectrum.

E_{n\nu} \approx E_{\mathrm{el}} + \left(\nu+\tfrac12\right)\hbar\omega + B\,J(J+1) \qquad (5)

where: $\hbar\omega$ — vibrational quantum, with $\omega=\sqrt{k/\mu}$ set by the curvature $k=\mathrm{d}^2E_n^{\mathrm{el}}/\mathrm{d}R^2$ at the minimum; $B=\hbar^2/(2\mu R_e^2)$ — rotational constant, with reduced mass $\mu$ and equilibrium separation $R_e$ ; $\nu,J$ — vibrational and rotational quantum numbers. Reading it: one clamped-nuclei calculation delivers $E_{\mathrm{el}}$ , $k$ , and $R_e$ , hence all three tiers. The approximation is not free. Because $\psi_n^{\mathrm{el}}$ depends on $R$ , the nuclear kinetic operator does not pass cleanly through the product — the Leibniz rule spills two extra derivative terms.

\hat T_N\!\left(\psi_n^{\mathrm{el}}\chi_{n\nu}\right) = \psi_n^{\mathrm{el}}\,\hat T_N\chi_{n\nu} - \sum_A\frac{\hbar^2}{M_A}\!\left(\nabla_A\psi_n^{\mathrm{el}}\right)\!\cdot\!\left(\nabla_A\chi_{n\nu}\right) - \sum_A\frac{\hbar^2}{2M_A}\!\left(\nabla_A^2\psi_n^{\mathrm{el}}\right)\!\chi_{n\nu} \qquad (6)

where: $\nabla_A$ — gradient with respect to nuclear coordinate $A$ ; the second and third terms are the nonadiabatic (derivative) couplings BO throws away. Reading it: BO keeps only the first term; the discarded pair each carry an explicit $1/M_A$ , which is why they are small. The size of that leading coupling is set by the electronic energy gap, via the Hellmann-Feynman theorem — the single most important formula for knowing when BO is trustworthy.

F_{nm}^{(1)}(R)=\langle\psi_n^{\mathrm{el}}|\nabla_A|\psi_m^{\mathrm{el}}\rangle = \frac{\langle\psi_n^{\mathrm{el}}|\nabla_A\hat H_{\mathrm{el}}|\psi_m^{\mathrm{el}}\rangle}{E_m^{\mathrm{el}}-E_n^{\mathrm{el}}}\quad(n\neq m) \qquad (7)

where: $F_{nm}^{(1)}$ — first-order derivative coupling between electronic states $n$ and $m$ ; the denominator $E_m^{\mathrm{el}}-E_n^{\mathrm{el}}$ is the electronic gap. Reading it: a large gap means small coupling and BO holds; a vanishing gap makes it diverge and BO collapses — the fault line of the second figure.

Limiting cases

The status of BO is a competition between the mass ratio, which weakens the coupling, and the electronic level spacing, which strengthens it when small.

Case	Result	Meaning
Well-separated electronic states	$\vert E_n^{\mathrm{el}}(R)-E_m^{\mathrm{el}}(R)\vert \gg \sqrt{m_e/M}\,E_{\mathrm{el}}$	Nonadiabatic coupling is negligible; the single-surface picture is excellent. This is the generic case for the ground state of a stable molecule.
Energy hierarchy	$E_{\mathrm{el}}:E_{\mathrm{vib}}:E_{\mathrm{rot}}\approx 1:(m_e/M)^{1/2}:(m_e/M)$	The natural separation of scales that justifies decoupling electronic, then vibrational, then rotational motion in turn.
Degeneracy / conical intersection	$E_n^{\mathrm{el}}(R)\to E_m^{\mathrm{el}}(R)$	Coupling diverges; BO fails. Electronic and nuclear motion mix, driving ultrafast internal conversion and photochemistry.
Heavy versus light nuclei	$m_e/M_{\mathrm H}\approx5.4\times10^{-4},\ \ m_e/M_{\mathrm I}\approx4.3\times10^{-6}$	BO improves with nuclear mass; hydrogen is the worst case, yet even there corrections are sub-percent away from crossings.

Common trapThe product factorization is exact, so the potential energy surface is a rigorous object

It is not exact. The factorization $\Psi\approx\psi^{\mathrm{el}}(r;R)\chi(R)$ deliberately drops the derivative couplings of equation (6), so a "potential energy surface" is an approximation-dependent construct, not a term in the true Hamiltonian. The neglected terms stay tiny only while the electronic states are well separated. Wherever two surfaces approach degeneracy — an avoided crossing or conical intersection — the coupling $F_{nm}^{(1)}\propto1/(E_m^{\mathrm{el}}-E_n^{\mathrm{el}})$ blows up and the surfaces mix. Vision, DNA photostability, and photosynthetic charge transfer all run through such regions, so never apply BO blindly near a near-degeneracy.

Core derivations

Derivation AThe energy-scale hierarchy from the mass ratio

The goal is to derive, from dimensional analysis and the single small parameter $m_e/M$ , the ordering $E_{\mathrm{el}}:E_{\mathrm{vib}}:E_{\mathrm{rot}}\sim1:(m_e/M)^{1/2}:(m_e/M)$ that sorts a molecular spectrum into the visible, infrared, and microwave. The argument needs only a molecular size $a$ , the electron mass $m_e$ , and a nuclear mass $M$ , and turns on the fact that the electronic physics setting the electronic scale also sets the bond stiffness.

Step 1. Estimate the electronic scale by confining an electron to a box of size $a$ : $E_{\mathrm{el}}\sim\hbar^2/(m_e a^2)$ , a few electron-volts for $a\sim1$ Å. Why: quantum confinement of the light electron sets the largest energy in the problem, the yardstick for the others.

Step 2. Vibrational scale $E_{\mathrm{vib}}\sim\hbar\omega$ with $\omega=\sqrt{k/M}$ , and fix $k$ from electronic physics: displacing the nuclei by $\sim a$ shifts $E_{\mathrm{el}}$ , so $k\sim E_{\mathrm{el}}/a^2\sim\hbar^2/(m_e a^4)$ . Why: the electrons supply the restoring force, so the bond stiffness is locked to the electronic scale, not independent.

Step 3. Substitute $k$ into $\hbar\omega$ : $\hbar\omega\sim\hbar\sqrt{\hbar^2/(m_e a^4 M)}=\hbar^2/(a^2\sqrt{m_e M})=E_{\mathrm{el}}\,(m_e/M)^{1/2}$ . Why: the mass enters as $\omega\propto1/\sqrt M$ and the stiffness as $1/\sqrt{m_e}$ , so the ratio to $E_{\mathrm{el}}$ is one power of $\sqrt{m_e/M}$ .

Step 4. Rotations from the nuclear moment of inertia $I\sim Ma^2$ : $E_{\mathrm{rot}}\sim\hbar^2/(2Ma^2)=E_{\mathrm{el}}\,(m_e/M)$ . Why: pure nuclear motion with no electronic stiffness carries the full power of $m_e/M$ , landing in the microwave.

Go deeper · Derivation B · 1 minFrom exact coupled equations to the Born-Oppenheimer approximation

Derivation BFrom exact coupled equations to the Born-Oppenheimer approximation

The goal is to place BO inside a controlled, exact framework so the approximation is a deliberate truncation rather than a guess. Expanding the true molecular eigenstate in the complete set of electronic states at each geometry yields exact coupled equations for the nuclear amplitudes; the adiabatic and Born-Oppenheimer approximations are then two successive levels of dropping the off-diagonal couplings.

Step 1. Write the exact ansatz $\Psi(r,R)=\sum_m\psi_m^{\mathrm{el}}(r;R)\,\chi_m(R)$ , a sum over the complete electronic basis with $R$ -dependent coefficients. Why: because $\{\psi_m^{\mathrm{el}}\}$ is complete at every fixed $R$ , this is exact — a change of representation, not an approximation.

Step 2. Insert into $\hat H\Psi=E\Psi$ and project onto $\psi_n^{\mathrm{el}}$ : $\left[\hat T_N+E_n^{\mathrm{el}}(R)\right]\chi_n-\sum_m\hat\Lambda_{nm}\chi_m=E\chi_n$ , with $\hat\Lambda_{nm}=\sum_A\frac{\hbar^2}{M_A}F_{nm}^{(1)}\!\cdot\nabla_A+\sum_A\frac{\hbar^2}{2M_A}G_{nm}$ . Why: $\hat H_{\mathrm{el}}$ puts $E_n^{\mathrm{el}}(R)$ on the diagonal, while $\hat T_N$ acting on the $R$ -dependent states generates the off-diagonal $\hat\Lambda_{nm}$ — the terms flagged in equation (6).

Step 3. Adiabatic approximation: keep the diagonal $\hat\Lambda_{nn}$ but drop all $n\neq m$ terms, giving $\left[\hat T_N+E_n^{\mathrm{el}}(R)-\hat\Lambda_{nn}\right]\chi_n=E\chi_n$ . Why: the diagonal term is a small shift keeping the nuclei on one surface; dropping the off-diagonals is justified when other states are remote.

Step 4. Born-Oppenheimer approximation: neglect $\hat\Lambda$ entirely, leaving $\left[\hat T_N+E_n^{\mathrm{el}}(R)\right]\chi_n=E\chi_n$ — equation (4). Why: the bare single-surface picture of spectroscopy, valid because $F_{nm}^{(1)}\propto1/(E_m^{\mathrm{el}}-E_n^{\mathrm{el}})$ is small when states are well separated.

Worked examples

1 · The three energy tiers of H₂ from the mass ratio (confidence builder)

From the electronic scale and the proton-to-electron mass ratio, estimate the vibrational and rotational quanta of $\mathrm{H}_2$ and compare with the measured $\hbar\omega=0.546$ eV and $B=7.5\times10^{-3}$ eV.

Step 1. Fix the inputs. The electron-to-proton ratio is $m_e/M=1/1836=5.45\times10^{-4}$ , a representative electronic scale is $E_{\mathrm{el}}\approx20$ eV (a couple of Rydbergs), and $(m_e/M)^{1/2}=0.0233$ .

Step 2. Vibrational quantum from $E_{\mathrm{vib}}\sim(m_e/M)^{1/2}E_{\mathrm{el}}$ : $E_{\mathrm{vib}}\approx0.0233\times20=0.47$ eV, against the measured $0.546$ eV — within 15 percent, an order-of-magnitude scaling law reproducing a real number.

Step 3. Rotational quantum from $E_{\mathrm{rot}}\sim(m_e/M)E_{\mathrm{el}}$ : $E_{\mathrm{rot}}\approx5.45\times10^{-4}\times20=0.011$ eV, against the measured $B=7.5\times10^{-3}$ eV — correct to within a factor of two.

Step 4. Read off the spectroscopy: $E_{\mathrm{el}}\sim$ eV sits in the visible-UV, $\hbar\omega=0.546$ eV gives $\lambda\approx2.3$ μm (infrared), and $B\sim10^{-2}$ eV lands in the far-infrared and microwave. The single parameter $m_e/M$ orders the whole $\mathrm{H}_2$ spectrum into $1:(m_e/M)^{1/2}:(m_e/M)\approx1:0.023:5\times10^{-4}$ .

2 · The HCl → DCl isotope shift, a direct test of BO

The vibrational wavenumber of $^{1}\mathrm{H}^{35}\mathrm{Cl}$ is $\omega_e=2990.9\ \mathrm{cm}^{-1}$ . Predict $^{2}\mathrm{D}^{35}\mathrm{Cl}$ using the BO claim that both isotopologues share one electronic curve, and compare with the measured $\approx2145\ \mathrm{cm}^{-1}$ .

Step 1. Invoke BO invariance. Because $E_n^{\mathrm{el}}(R)$ is computed with the nuclei clamped, it is mass-independent, so HCl and DCl move in the identical well with the same force constant $k$ . Only the reduced mass in $\omega=\sqrt{k/\mu}$ changes: $\omega_{\mathrm{DCl}}/\omega_{\mathrm{HCl}}=\sqrt{\mu_{\mathrm{HCl}}/\mu_{\mathrm{DCl}}}$ .

Step 2. From the atomic masses $m_{\mathrm H}=1.00783$ , $m_{\mathrm D}=2.01410$ , $m_{\mathrm{Cl}}=34.9689$ u: $\mu_{\mathrm{HCl}}=0.9796$ u and $\mu_{\mathrm{DCl}}=1.9044$ u.

Step 3. Form the ratio $\mu_{\mathrm{HCl}}/\mu_{\mathrm{DCl}}=0.5144$ , square root $0.7172$ , so $\omega_{\mathrm{DCl}}=2990.9\times0.7172=\mathbf{2145\ cm^{-1}}$ .

Step 4. The prediction matches the observed DCl fundamental to four figures. This near-exact agreement is a clean confirmation of BO: the potential curve really is mass-independent, and the whole isotope shift comes from the reduced mass alone.

3 · How good is BO — hydrogen versus iodine

Quantify the BO expansion parameter $\kappa=(m_e/M)^{1/4}$ and the size of post-BO energy corrections for the lightest nucleus (hydrogen) and a heavy one (iodine-127).

Step 1. Hydrogen: $m_e/M_{\mathrm H}=1/1836=5.45\times10^{-4}$ , so $(m_e/M)^{1/2}=0.0233$ and the Born-Huang parameter $\kappa=(m_e/M)^{1/4}=0.153$ .

Step 2. Iodine: the $^{127}\mathrm{I}$ mass is $126.90\ \mathrm{u}=2.31\times10^{5}\,m_e$ , so $m_e/M_{\mathrm I}=4.32\times10^{-6}$ , giving $(m_e/M)^{1/2}=2.08\times10^{-3}$ and $\kappa=0.046$ .

Step 3. Leading nonadiabatic corrections scale as $\kappa^4\sim m_e/M$ : about $5\times10^{-4}$ for hydrogen, $4\times10^{-6}$ for iodine (away from any near-degeneracy). So even hydrogen, the worst case, carries only a 2 percent velocity-scale separation and sub-percent energy corrections; for heavy nuclei BO is accurate to parts per million — a genuinely small-parameter expansion.

Physics of the portrait

The Born-Oppenheimer approximation is the moment quantum mechanics stopped being a theory of atoms and became a theory of molecules — worked out at Göttingen in 1927, one year after Schrödinger's equation, by the man who gave quantum mechanics its statistical interpretation and his young doctoral student.

Origins · 1 minMax Born and how this was discovered

Max Born

1882–1970 · Göttingen quantum mechanics and the theory of molecules

Max Born ran the Göttingen institute that, alongside Copenhagen, was the crucible of quantum mechanics: with Heisenberg and Jordan he cast the new mechanics in matrix form in 1925, and in 1926 gave the wavefunction its probabilistic meaning — the interpretation that won him the 1954 Nobel Prize. In 1927, with his doctoral student J. Robert Oppenheimer, Born published Zur Quantentheorie der Molekeln, showing that the small mass ratio lets the molecular wavefunction be expanded in powers of $\kappa=(m_e/M)^{1/4}$ : at leading order electrons sit on a fixed nuclear frame and the nuclei then move in the resulting electronic energy. The paper turned an intractable many-body equation into the two-step scheme still used today, and quietly created the central object of chemical physics — the potential energy surface. Born returned to it in 1954 with Kun Huang, giving the systematic expansion that bears both names. The approximation is now so foundational that "the potential energy surface" is spoken of as though physically real, when it is in fact Born and Oppenheimer's approximation made concrete — an idealization whose breakdown at conical intersections is itself a thriving field.

History

1927

— Born and Oppenheimer publish "Zur Quantentheorie der Molekeln," separating electronic from nuclear motion

1929

— von Neumann and Wigner prove the non-crossing rule for electronic states of a diatomic

1937

— Teller shows that surfaces of polyatomic molecules generically touch at conical intersections

1954

— Born and Huang give the systematic expansion of molecular dynamics in the electron-to-nuclear mass ratio

1963

— Herzberg and Longuet-Higgins find the geometric-phase sign change of the wavefunction around a conical intersection

Going further

Texts. Foot, Atomic Physics, chapter on molecular structure · Demtröder, Laser Spectroscopy, molecular-spectra chapters · Atkins and Friedman, Molecular Quantum Mechanics, chapter on the Born-Oppenheimer separation · Born and Huang, Dynamical Theory of Crystal Lattices (1954), the systematic expansion · Born and Oppenheimer, Ann. Phys. 84, 457 (1927), the original paper.

Provenance. The Hamiltonian, clamped-nuclei equation, product ansatz, and coupled-equation derivation follow Born and Oppenheimer (1927) and standard treatments in Demtröder and Foot. A verifier should confirm: the hierarchy $1:(m_e/M)^{1/2}:(m_e/M)$ from the dimensional argument; the mass ratios $m_e/M_{\mathrm H}\approx5.4\times10^{-4}$ and $m_e/M_{\mathrm I}\approx4.3\times10^{-6}$ (from $M_{\mathrm H}\approx1836\,m_e$ , $1\ \mathrm{u}=1822.9\,m_e$ ); the HCl → DCl shift $\omega_{\mathrm{DCl}}=\omega_{\mathrm{HCl}}\sqrt{\mu_{\mathrm{HCl}}/\mu_{\mathrm{DCl}}}=2145\ \mathrm{cm}^{-1}$ ; the Hellmann-Feynman form of $F_{nm}^{(1)}$ ; and the H₂ anchors $\hbar\omega=0.546$ eV, $B=7.5\times10^{-3}$ eV. No fabricated constants are used.

Prerequisites. The Chemical Bond and Molecular Potential Energy Curves · Molecular Orbital Theory and Bonding/Antibonding States. Leads to. The Harmonic Oscillator Model of Vibration · The Rigid Rotor Model · The Franck-Condon Principle · Anharmonicity and the Morse Potential.

Ask the tutor: Ask me to trace how the couplings of equation (6) blow up at a conical intersection, or to work the BO energy scales for any diatomic.

Is the Born-Oppenheimer approximation the same as saying the nuclei do not move?

No — that is a common oversimplification. BO does not freeze the nuclei; it separates the two motions by timescale. The nuclei absolutely move, and their vibration and rotation are the whole content of equation (4). What BO assumes is that the electrons relax to a given state at every instant, so the electronic wavefunction depends only on the current geometry, not on how fast the nuclei pass through it. You clamp the nuclei only during the electronic calculation, then release them to move on the surface.

What is the difference between the "adiabatic" and "Born-Oppenheimer" approximations?

They are two successive levels of neglect in the coupled equations of Derivation 2. The adiabatic approximation keeps the diagonal coupling $\hat\Lambda_{nn}$ , a small mass-dependent shift, while dropping all off-diagonal terms — the nuclei stay on a single, slightly corrected surface. The Born-Oppenheimer approximation drops even the diagonal term, so the nuclei move on the bare eigenvalue $E_n^{\mathrm{el}}(R)$ . The diagonal correction is tiny (of order $m_e/M$ ), so the two agree excellently; the distinction matters only in precision spectroscopy of light molecules.

Why does a small electronic energy gap make the approximation fail?

The nonadiabatic coupling of equation (7) carries the gap $E_m^{\mathrm{el}}-E_n^{\mathrm{el}}$ in its denominator. Physically, the coupling measures how much the electronic wavefunction is reshaped as the nuclei shift; when two surfaces are close, a small displacement can flip the character of the state from one surface to the other, so $\psi^{\mathrm{el}}$ changes violently with $R$ and $\nabla_A\psi^{\mathrm{el}}$ becomes large. As the gap closes at a conical intersection the coupling diverges, the surfaces mix completely, and there is no single surface left for the nuclei — internal conversion becomes ultrafast.

If the potential energy surface is only an approximation, why is it treated as real?

Because it is spectacularly useful and, away from crossings, spectacularly accurate — corrections are parts per thousand for hydrogen and parts per million for heavy atoms (Worked Example 3). Equilibrium geometries, dissociation energies, vibrational frequencies, and reaction barriers are all defined on the BO surface and agree with experiment, so chemists reason about molecules as if the surface were a physical landscape. The caveat is honesty: the surface appears only after the derivative couplings are discarded, and it loses meaning exactly where those couplings matter. "Real" here means "an excellent effective description," not "a term in the exact Hamiltonian."

How does BO give rise to the Franck-Condon principle?

The Franck-Condon principle is BO applied to an electronic transition. Because electrons move so much faster than nuclei, an excitation happens while the nuclei are effectively frozen — a "vertical" jump between two BO surfaces at fixed $R$ . The nuclear wavefunction $\chi_{n\nu}$ has no time to change during the jump, so a vibronic band's intensity is set by the overlap $|\langle\chi_{n'\nu'}|\chi_{n\nu}\rangle|^2$ on the two surfaces — the Franck-Condon factor. Without the BO separation there would be no distinct surfaces to jump between and no overlap to control the intensities.

Does BO break down more easily for lighter or heavier molecules?

Lighter. The whole approximation is an expansion in $m_e/M$ , so corrections shrink as the nuclei get heavier — iodine beats hydrogen by two orders of magnitude (Worked Example 3), and hydrogen-containing molecules show the largest nonadiabatic effects and isotope shifts. But mass is only half the story: any molecule, however heavy, breaks BO catastrophically at a geometry where two surfaces become degenerate — there the gap, not the mass, controls the failure.